[Bug] [Windows] Error: "Expression: cannot seek vector iterator after end" when running mmdeploy::Detector::Apply in debug version of mmdeploy library.

[antoszy]

Checklist

• I have searched related issues but cannot get the expected help.
• 2. I have read the FAQ documentation but cannot get the expected help.
• 3. The bug has not been fixed in the latest version.

Describe the bug

When running mmdeploy::Detector::Apply function I get the following error:

Reproduction

Build debug version of the mmdeploy libraries with:

cmake .. -G "Visual Studio 16 2019" -A x64 -T v142 `
   -DMMDEPLOY_BUILD_SDK=ON `
   -DMMDEPLOY_BUILD_EXAMPLES=ON `
   -DMMDEPLOY_BUILD_SDK_PYTHON_API=ON `
   -DMMDEPLOY_BUILD_SDK_CXX_API=ON `
   -DMMDEPLOY_TARGET_DEVICES="cuda" `
   -DMMDEPLOY_TARGET_BACKENDS="trt" `
   -Dpplcv_DIR="$env:PPLCV_DIR/build_d/install/lib/cmake/ppl" `
   -DTENSORRT_DIR="$env:TENSORRT_DIR" `
   -DCUDNN_DIR="$env:CUDNN_DIR" `
   -Wno-dev `
   -DOpenCV_DIR="$env:OPENCV_DEBUG_DIR\..\.."
cmake --build . --config Debug -- /m
cmake --install . --config Debug

Crate TensorRT version of the model by running the following command:

python mmdeploy\tools\deploy.py \
	.\mmdeploy\configs\mmdet\instance-seg\instance-seg_tensorrt_static-800x1344.py  \
	.\mmdetection\configs\mask_rcnn\mask_rcnn_r50_fpn_1x_coco.py  \
	.\mmmodels\mask_rcnn_r50_fpn_2x_coco_bbox_mAP-0.392__segm_mAP-0.354_20200505_003907-3e542a40.pth  \
	.\mmdetection\demo\demo.jpg  \
	--work-dir .\mmdeploy_model\mask-rcnn  \
	--device cuda  \
	--dump-info \

Create a simple project in Visual Studio 2019 with the debug mmdeply library with the following code:

mmdeploy::Model _model("model_path");
mmdeploy::Detector = _detector(_model, mmdeploy::Device{ "cuda", 0 });
auto dets = _detector.Apply(cv_mat_img);

Compile, link with debug mmdeploy libraries and run the application.

Environment

- Windows
- IDE: Visual studio 2019 with v142 build tools
- Nvidia graphics card with CUDA
- MMDepoly version: v0.11.0

Error traceback

The error comes from this line:

mmdeploy/csrc/mmdeploy/backend_ops/tensorrt/multi_level_roi_align/trt_multi_level_roi_align.cpp

Line 83 in 0b6b9ee

std::vector<float>(mFeatmapStrides.begin(), mFeatmapStrides.begin() + nbInputs - 1);

I've identified the problem and I will create a PR to fix it.

[mbed92]

The same for the newest Visual Studio 2022 (v143)